The World Bank MAY NUMIBER 23 PUBLIC SECTOR Using surveys for public 2 1 558 sector reform As Uganda shows, well-designed surveys can give policymakers the information they need to reallocate funds and improve the delivery of services. Data that can be used to inform policy deci- Uganda shows that a survey can provide a sions are typically scarce in low-income coun- useful check on the supply side of service tries, where standard policy prescriptions delivery when institutions perform poorly are less likely to apply. Interventions based and official statistics are lacking or of poor on inadequate information and thus mis- quality. Diagnostic surveys guided assumptions may not achieve Expenditure tracking surveys began in expected results-despite the fact that sub- Uganda once it was realized that, while can provide vital stantial public or donor funds are being public spending on basic services had spent. For example, an adjustment opera- increased substantiallysince the late 1980s information when tion that focuses on spending allocations (albeit from a small base), several officially may achieve its benchmarks but have no reported outcome and output indicators institutional effect on actual service delivery. remained stagnant. World Bank structural Diagnostic surveys can provide vital infor- adjustment programs specifically sup- weaknesses inhibit mation for decisionrnakers when institu- ported increased spending on basic ser- tional weaknesses inhibit a more regular vices and contained provisions to protect a more regular flow flow of information. If strategically designed, budget allocations. The most obvious dis- a survey can help induce policy change by parity in outcome indicators was seen in of information pointing directly to the main bottlenecks, primary school enrollments. Budget allo- making it easier for policymakers to find cations for primary education nearly tripled solutions. The note summarizes two cases between 1991 and 1995-yet there was in Uganda where diagnostic sur veys proved hardly any increase in officially-reported particularly useful. The first case involves enrollment. public spending on health and education; The hypothesis was that actual service the second considers tax administration delivery (output) was much worse than bud- from the perspective of taxpaying firms. get allocations would imply because pub- lic funds (inputs) did not reach the intended Surveying service delivery facilities as expected. Reasons for facili- When assessing service delivery, most sur- ties not receiving the allocated funds could veys examine the effects of policies or inter- range from competing priorities at vari- ventions on households and their demands ous levels of government to corruption and for and perceptions of the quality of ser- misuse of public funds. To test this hypoth- vices. Inputs and outputs on the supply esis, budget allocations and actual spend- side-such as the flow of public funds and ing were compared in two decentralized school enrollment rates-are left for offi- sectors, primaryeducation and health care. cial statistics or administrative records. Because local government accounts were FROM THE DEVELOPMENT ECONOMICS VICE PRESIDENCY AND POVERTY REDUCTION AND ECONOMIC MANAGEMENT NETWORK not generally available, a field survey was Relative to nonsalary expenditures, track- carried out in 19 of Uganda's 39 districts, ing teacher salaries was complicated by the covering a randomly selected sample of 250 absence of disaggregated central govern- government schools and 100 public health ment pay data. But salaries seem to have clinics, to collect spending data for 1991-95. reached schools much better than nonwage The survey was designed and implemented allocations-thoughwith considerable delay. in collaboration with the Ugandan gov- The only systematic way of misappropriat- ernment, a local research center (Economic ing salary funds was through "ghost" teach- Policy Research Centre) and an indepen- ers on the payroll. Previous efforts by the dent Ugandan consulting firm (MSE Con- government to clean up the teachers' pay- Official statistics sultants). roll give some idea of the magnitude of the leakage in salaries: in 1993 nearly 20 per- cannot always be Findings on primary education cent of the ghost teachers on the payroll The field survey confirmed the hypothe- were removed. trusted sis that input flows suffer from serious prob- The survey also unearthed a host of other lems that largely derive from weak features of the service delivery system that governance and lack of accountability. On were unexpected but crucial to the func- average, less than 30 percent of the funds tioning of health and education services- intended for nonsalary public spending and hence to any intervention in these actually reached schools in 1991-95 (fig- sectors. First, in contrast to the stagnation ure 1), because district authorities kept in enrollments indicated by officially and used most of the nonsalary capita- reported data, the school surveyfoundthat tion (perstudent) grants meant for schools. primary enrollment actually increased 60 Similarly, schools were allowed to keep, percent in 1991-95. Such a stunning dis- at best, only a third of mandatory tuition crepancy indicates that official statistics can- fees from parents. The rest went to district not always be trusted. education offices. There were also large Second, although the government's share variations: at the median, school retention of spending on public primary education of both capitation grants and tuition fees increased over time, most of the burden con- was zero. tinued to be borne by parents, who accounted for as much as 70 percent of FIGURE 1 AVERAGE CAPITATION GRANT PER STUDENT IN UGANDA, 1991 PRICES school spending in 1991 and 60 percent US. dollarsperstud1t in 1995 (40 percent and 20 percent at the Intended grant median school, respectively). Despite higher amount public spending, parents' contributions con- 3.0 tinued to increase in real terms over the sur- vey period. 2.5 Third, the survey showed that the per- 2.0 formance of public facilities in different sec- tors can vary considerably even within one 1.5 country, depending on the facilities' insti- 1.0 tutional context and incentives. In Uganda Recieved schools keep systematic records of financial .5 , _ b _chol flows and enrollments. Health units and (mean) local governments (districts), by contrast, t991 1992 1993 1994 1995 do not keep good records. Minimum: 0 Minimum: 0 Minimum: 0 Minimum: 0 Minimum: 0 Median: 0 Median: 0 Median: 0 Median: 0 Median: 0 Findin on rima health care Maximum: 0.8 Maximum: 0.9 Maximum: 1.3 Maximum: 1.3 Maximum: 1.0 gS pr ry Aote: The 1991-92 exchange rate was used for all years. Minimum, median, and maximum refer to funds The health survey found that clinic records recieved by schools. of resource flovs were kept in kind-if at Smnece: Ablo and Reinikka 1998. all-and that there were essentially no PREMNOTE 23 records for clinic inpatients or outpatients. tricts. This experience demonstrates that Interviews at the clinics revealed that most data can be a powerful tool of change. In medical supplies reached health facilities. health care, where the survey could not pro- This was because the supplies did not go vide clear-cut findings for policyrnakers to through local governments but were dis- act on, improvements in accountability have tributed to health units directly from the been less evident. center. For the most part, however, the sur- vey instrument that had worked well in Findings on tax administration schools was unable to provide substantive Most analyses of taxation in low-income quantitative information for health units. countries focus on revenue and the formal From observations, discussions with focus tax structure, and fail to address adminis- Data can be a groups, and other qualitative research meth- trative realities. But if it is not managed fairly ods, researchers learned that many problems and efficiently, tax administration can cre- powerful tool of originated within clinics, where staff expro- ate major distortions-no matter how well priated a large portion of drugs and med- designed a tax system is. It is often argued change ical supplies and sold them onward. Such that administrative practices and tax evasion leakage was estimated to average nearly 70 cannot be measured because information percent.Asaresultthepoor,whocannotafford is confidential. The experience in Uganda these misappropriated supplies, often do not points to the opposite. receive even basic health services. These results A survey of 243 randomly selected firms indicate that the health system continues to operating in Uganda was carried out in 1998 suffer the legacy of the oppressive state in to obtain information on a range of their busi- the 1970s and early 1980s and that past sur- ness activities, including a diagnostic sur- vival mechanisms of health staff (whose salaries vey on tax administration. With appropriate were extremely low) have not changed. survey questions and interview techniques, respondents-who were assured of strict con- Policy changes fidentiality-were willing to provide both Although many of the survey findings were qualitative information and quantitative data troubling, there is a positive side to the story. on sensitive issues, such as taxation. After the survey results were released in 1996, the government responded actively to them. Firms 'perceptions For example, to increase transparency and The survey allowed an examination of accounLability, iiontlhly transfers of public changes in firimis' perceptions over time, funds to districts are now reported in the using rankings and numerical scores. Firms main newspapers and broadcast on radio. reported that tax evasion, which was found Moreover, government policy requires that to be the main reason for unfair competi- transfers to primary education be displayed tion in a similar survey in 1994, continues on public notice boards in each school and to be perceived as a leading constraint to district center, and the Ministry of Educa- competition. But the numerical scores for tion monitors compliance. competitors evading taxes have declined. School-based procurement has replaced Despite improving perceptions, poor com- the central supply of construction and other pliance continues. Manufacturing firrns esti- materials. Detailed data on spending on mated that half of their competitors gain an teacher salaries are now available at the cen- advantage through tax evasion. In com- tral government level. The Ministry of Edu- mercial agriculture, where the share of tax- cation replicated the school survey in 1998 paying firms is the lowest, only 5 percent and found major improvements in the flow of competitors were perceived to evade taxes. of funds. Finally, an effort is being made- with support from the International Devel- Quantitative data opment Association-to institute basic The survey also provided a wealth of quan- public accounting systems that include dis- titative information that can be used to exam- MAY 1999 ine the extent to which firms have access But it is already evident that issues relating to tax exemptions, disagree with the revenue to taxation have become more prominent authority about their tax assessments, and in the policy dialogue between the govern- are subjected to tax audits. In addition, these ment and the Ugandan private sector. data can determine which firm character- istics are correlated with tax audits, exemp- Further reading tions, and the like. Quantitative information Ablo, Emmanuel, and Ritva Reinikka. 1998. also covers corruption in tax administration. "Do Budgets Really Matter? Evidence For example, one in four of the audited firms from Public Spending on Education and reported incurring additional costs such Health in Uganda." Policy Research Work- as bribes. Interestingly, firms whose own tax ing Paper 1926. World Bank, Washing- Survey data can assessmentdiffered by 100 percent or more ton, D.C. from that of the revenue authority reported Asiimwe, Delius, Francis Mwesigye, Barbara determine which that they always had to paybribes to tax offil- McPake, and Pieter Streefland. 1996. cers-while average survey firms reported "Informal Health Markets and Formal firm characteristics that they were required to pay bribes only Health Financing Policy in Uganda." seldomly. Makerere Institute of Social Research, are correlated The survey also highlighted implemen- Ministry of Health (Uganda), London tation problems associated with the 1996 School of Hygiene and Tropical Medi- with tax audits, introduction of the value added tax. In par- cine, and Royal Tropical Institute (Ams- ticular, the survey has proven useful for terdam). exemptions, and assessing compliance with the tax and for Chen Duanjie, and Ritva Reinikka. Forth- ensuring that the cost of doing business has coming. "Can Taxation Be a Constraint the like not been increased by bad administrative is a Low Revenue Economy? Evidence practices. For example, 42 percent of the from Uganda. "Africa Region Working surveyed firms that applied for an input tax Paper Series. World Bank, Washington, refund received their expected amount. D.C. A significant portion (18 percent) of firms Republic of Uganda. 1998. "Monitoring and that claimed an input tax credit did not Evaluation Accountability and Trans- receive any refund, while the rest (40 per- parency of Schools and Districts for UPE cent) received a partial refund. Further- Funds." Ministry of Education and Sports, more, the waiting period for a full or partial Kampala. refund can be lengthy. Among firms that received at least part of the refund, more This note was written by Ritva Reinikka (Senior than 50 percent waited for more than six Country Economist, Macroeconomics 2, Africa weeks-while 10 percent waited for more Region). than six months. Ifyou are interested in similar topics, consider The data provided by the firm survey can joining the Public Expenditure Thematic Group be used to help the Ugandan government (contact Vinaya Swaroop, x37688, or Malcolm tackle problems of tax administration- Holmes, x37189) or the Tax Policy and Admin- including corruption-more effectively. This istration Thematic Group (contact Luca Barbone, survey was completed only recently, and x32556, orRonnieDas-Gupta, x87764), orclick its full impact on policy remains to be seen. on Thematic Groups on PREMnet. This note series is intended to summarize good practice and key policy findings on PREM-related topics. PREMnotes are distributed widely to Bank staff and are PIE h also available on the PREM website (http://prem). If you are interested in writing a PREMnote, email your idea to Sarah Nedolast. For additional copies of this PREM- Pm#Siubs note please contact the PREM Advisory Service at x87736. Prepared for World Bank staff